Our MIART 2026 paper DoMa-Seg, a collaboration led by PhD student Jihe Li, builds radiation dose into segmentation training: 12% lower dosimetric error for target volumes and 1.8× faster training than a state-of-the-art dose-aware method. Code.